Exploring changes to support Plasma (beam machines)

[James Walker]

Hi everyone.  I am exploring changes to the code base to scratch my plasma cnc itch.  :-)

I have forked this code to my github repo to do this work:

https://github.com/joco-nz/dxf2gcode-plasma

I'm not sure what the protocols are re providing work back but obviously anything I do can be pulled and merged/polished/improved back into the main code line.

I may lean things up to be a VERY focused version on beam machine needs only just to keep the code line I am working on a little tighter. But we shall see.

Anyway, great tool and please grab any of the hacking I do to the benefit of all.

As a side note, I started a question conversation in the users group and Christian replied which I appreciate.  BUT all my follow up responses are getting auto deleted!

So if I appear to post once and then go dark it is because google groups has it in for me and my follow up messages are being killed off like in the users group.

Cheers - James.

[Christian Kohlöffel]

Hi James,

 

That’s really strange. I just checked for spam e-mails or similar, but there is nothing reaching the server. Just continue here, it seems to work fine J …

 

regards

Christian

--
--
You received this message because you subscribed to the Google
Groups-group "dxf2gcode-dev".
To post a message, send mail to dxf2gc...@googlegroups.com
To unsubscribe, send mail to dxf2gcode-de...@googlegroups.com
See http://groups.google.de/group/dxf2gcode-dev?hl=en for more options
and the dxf2gcode project page at http://code.google.com/p/dxf2gcode/
---
You received this message because you are subscribed to the Google Groups "dxf2gcode-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dxf2gcode-de...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dxf2gcode-dev/55482987-7f7d-4b64-90b6-00a9d7b5533an%40googlegroups.com.

[James Walker]

Will do.

Having fun at the moment. Have working lead-ins and lead-outs using the CAM side path compensation.  Have added some flags to the layer processing to allow mass adjustment to the style of leads and suppression of either leadin or leadout or both.

Have encountered the dreaded "No intersection" warning.  I don't have issues with geom I make but some of the dxf files I have picked off the net are breaking on the simplest shapes due to screwy geom hidden in the lines.

Is there anyway to "clean" some of this on load? Or with another tool before hand?  In many cases it seems like some verts just need to be welded together based on a proximity circle.  I realise from reading that is not the whole issue but I know it is part of it.  As I recreated a simple shape from scratch that was breaking and it worked exactly as expected. I can supply the offending dxf if that helps.  No to include in this message conversation from what I can see.

Cheers - J.

[Christian Kohlöffel]

Hi James,

 

i have tried to add some Kind of preprocessing in the shapeoffset.py file to fix such „bad“ dxfs. However i didn’t fully finished and it only helps for lines i guess. The Problem still exists for arcs’s i would guess

Feel free to check and improve.

Regards

To view this discussion on the web visit https://groups.google.com/d/msgid/dxf2gcode-dev/08ddda06-e74b-4707-91e9-042f6c924a09n%40googlegroups.com.

 

[James Walker]

Christian - any hints on where to look?  e.g. function/general line range? I'm not that familiar with that code area so a point on where to start looking could help a lot.

Probably also a sanity check on how dxf line segments work is in order.  The have a start (Ps) and end (Pe) point.  I assume that when constructing a shape of lines the lines have an order on how they are linked. I have seen code you have to convert those to CW as base standard.

If we consider of a triangle it could be something like:

Line 1:   Ps:100,0 Pe:0,0

Line 2:   Ps:0,0 Pe:50,100

Line 3:   Ps: 50,100 Pe: 100,0

What I think I am seeing when there is a problem is lines that are more like:

Line 1:   Ps:100,0 Pe:0,0

Line 2:   Ps:0,0 Pe:50.025,100

Line 3:   Ps: 50,100 Pe: 100,0

From this definition line 2 and line 3 are not actually linked and closing the shape by a fractional amount. The "no intersection warning" seems to be linked to line segments that have these fractional miss-alignments on the point that should be the join between them.

If I am off the mark just let me know.

Cheers - J.

[James Walker]

I have been looking into this more and I believe I was completely off target in my last message.   After some investigation I believe one of the issues I encounter from dxfs that are not well created is per attached image. This Crossing Line problem breaks the CAM side cutter compensation.  I have a sample method that cleans for this problem as part of the shape loading process.   This extra line segment is extremely small (sub 1mm) so on screen is not visible until zoom in a massize amount.  Also hunting for these in a cad package such as LibreCAD seems to be messy.  Initial testing of my cleaning method is promising.

Cheers - J.

[Christian Kohlöffel]

Hi James,

 

this is a good example of badly created dxf files. And there are many examples out there I believe. Thanks for digging into this, that will bring the compensation algorithm a big step ahead

 

just tell me when it will be worth testing, I can for sure dig out some more bad examples J ….

 

Regards

Christian

To view this discussion on the web visit https://groups.google.com/d/msgid/dxf2gcode-dev/1a60479d-9270-43fb-92f8-9c89f9ba70abn%40googlegroups.com.

[James Walker]

Christian  - if you have a few example dxf's I can use for some checks that would be good.  Maybe add them to what I see as the upstream repo in side the dxf test folder, perhaps as a sub folder called "poordxf" or something else indicating they are badly formed.

That would allow me to do a few more sanity checks.

The tests I have been doing seems promising.  But I'm sure I will have missed something.

Cheers - J.

[Christian Kohlöffel]

Here one example from a ticket:

dxf2gcode / Tickets / #126 Why "left compensation" works only on switched direction (sourceforge.net)

added to the folder.

 

I added on other the folder too.

 

Commiting in development.

 

Regards

Christian

 

 

Von: 'James Walker' via dxf2gcode-dev [mailto:dxf2gc...@googlegroups.com]
Gesendet: Mittwoch, 19. Mai 2021 08:25
An: dxf2gcode-dev <dxf2gc...@googlegroups.com>
Betreff: Re: [dxf2gcode-dev] Exploring changes to support Plasma (beam machines)

 

Christian  - if you have a few example dxf's I can use for some checks that would be good.  Maybe add them to what I see as the upstream repo in side the dxf test folder, perhaps as a sub folder called "poordxf" or something else indicating they are badly formed.

 

That would allow me to do a few more sanity checks.

 

The tests I have been doing seems promising.  But I'm sure I will have missed something.

 

Cheers - J.

 

On Wednesday, May 19, 2021 at 5:37:12 PM UTC+12 Chrisko wrote:

Hi James,

 

this is a good example of badly created dxf files. And there are many examples out there I believe. Thanks for digging into this, that will bring the compensation algorithm a big step ahead

 

just tell me when it will be worth testing, I can for sure dig out some more bad examples J ….

 

Regards

Christian

 

Von: 'James Walker' via dxf2gcode-dev [mailto:dxf2gc...@googlegroups.com]
Gesendet: Mittwoch, 19. Mai 2021 00:35
An: dxf2gcode-dev <dxf2gc...@googlegroups.com>
Betreff: Re: [dxf2gcode-dev] Exploring changes to support Plasma (beam machines)

 

I have been looking into this more and I believe I was completely off target in my last message.   After some investigation I believe one of the issues I encounter from dxfs that are not well created is per attached image. This Crossing Line problem breaks the CAM side cutter compensation.  I have a sample method that cleans for this problem as part of the shape loading process.   This extra line segment is extremely small (sub 1mm) so on screen is not visible until zoom in a massize amount.  Also hunting for these in a cad package such as LibreCAD seems to be messy.  Initial testing of my cleaning method is promising.

 

Cheers - J.

 

To view this discussion on the web visit https://groups.google.com/d/msgid/dxf2gcode-dev/e13afafe-d4d6-4027-8ec6-ebcfec610c06n%40googlegroups.com.